Review our method
Reproduce the Skolkoll score from the specification, fixtures and script, then submit your own review.
Quick check
Score v1, weights, thresholds, formulas and a worked example that should produce 74.
The package is open on GitHub; serious methodological reviews are published with responses and corrections.
Specification, synthetic fixture, real extract and reproduction harness point to current main.
1. An invitation to scrutinise
The Skolkoll score is a composite measure built from public data from Skolverket, SCB and Kolada. A measure like that is only as credible as its method is open. So we have made the whole calculation reviewable: every weight, threshold and formula is documented and cited to a line in the code, and we have packaged everything needed to recompute the score from scratch — with no access to our server.
An external statistician should be able to reproduce the score from this package alone. Theworked example in the specification starts from five dimensions ([80, 70, 67, 85, 60] at the weights [0.30, 0.25, 0.20, 0.15, 0.10]) and arrives at aSkolkoll score of 74 — the same value that is verified live in the test suite. If you get a different number, either we or you have found something, and we want to hear about it.
The review package
The merged, version-tagged artefacts are openly available on GitHub. The links below point to the current version in main; the specification is version-tagged (v1 /scoreVersion 1.0.0), and to reproduce an exact historical score version use the matching tagged release.
- Specification: the Skolkoll score v1 — the full methodology (
scoreVersion 1.0.0) with every weight, threshold and formula, each cited to a line in the scoring code and, where a golden value exists, to the parity test. - Fixture: synthetic pool — the primary reproducible fixture. Fully invented schools (A–E) and statistics, so there is no privacy risk; intended to be fed straight into the reproduction harness.
- Fixture: real extract — a small, privacy-safe sample drawn from the public dataset (school-level statistics with no personal data), so you can check the method against real numbers and not just against constructed cases.
- Reproduction harness — a standalone Node script that recomputes the Skolkoll score from the fixtures above, so you can verify the number end-to-end on your own machine.
How the underlying numbers are calculated is documented openly on the method page. This page is about how you can check us.
2. What we commit to
A commitment that can't hurt isn't credible. So this is a genuine commitment, not a marketing line:
We publish methodological reviews and critiques in full — including ones that are critical of us— without censoring them.
We want to be honest about what this is and is not. There is no automated feature in the product that publishes incoming reviews. This is an editorial commitment: when we receive a serious review, we undertake to publish it — for example on a method-review page or in thechangelog — together with our response and any corrections. So we are describing a promise we intend to keep, not a system behaviour that is already built. If a review shows that the score is wrong, we fix it and document the change openly.
3. How to submit a review
Send your review, question or objection toinfo@skolkoll.se. This is the same channel through which we already invite reviewers in themethod policy. Please say which part of the method it concerns — a particular dimension, weight, threshold or a line in the specification — so the reply can be more concrete. You can also reach us via the support page.